Would somebody agree that a MySQL/PostgreSQL cluster has more downtime than a replicated cluster? How often do high availability mechanism cause downtime itself?